[] []

Hardware

One development board and one expansion board, used as a pair. No additional modules and no soldering are required.

What is on the board

ESP32-S3 N16R8: 16 MB of flash and 8 MB of octal PSRAM. The audio slots are allocated in PSRAM.
ES8388: an audio codec with both an ADC and a DAC, configured over I2C and carrying samples over I2S.
HT6872: a class-D amplifier wired bridge-tied, powered from 5 V, its CTRL pin enabled through a voltage divider.
Two GMI6027P-56DB electret microphones, on the codec’s LIN1 and RIN1 inputs.
A speaker on connector J1, driven by the amplifier.

Pin map

Taken from the firmware. The manufacturer’s schematic names the same signals from the codec’s point of view, so the data pins read reversed there; see the codec page.

Signal	GPIO	Function
SW1	12	button 1; hold to record
SW2	11	button 2
SW3	13	button 3
LED	21	blinks while a slot is chosen
I2C SDA	14	ES8388 control
I2C SCL	47	ES8388 control
I2S MCLK	8	codec master clock
I2S BCLK	3	bit clock
I2S LRCLK	9	word clock
I2S DOUT	46	ESP32 → ES8388 DSDIN (DAC)
I2S DIN	10	ES8388 ASDOUT → ESP32 (ADC)

play_tone/main/main.c

#define I2C_SDA_PIN         14
#define I2C_SCL_PIN         47

#define I2S_MCLK_PIN        8
#define I2S_BCLK_PIN        3
#define I2S_LRCLK_PIN       9
#define I2S_DOUT_PIN        46
#define I2S_DIN_PIN         10

Buttons and LED

Each button sits between its GPIO and ground, pulled up on the board through 100 kΩ, with a 100 nF capacitor across it. A pressed button therefore reads low. The firmware also enables the ESP32’s internal pull-up, so the input is defined on boards populated without the external resistor.

Button wiring, identical for SW1, SW2 and SW3. The capacitor filters mechanical bounce in hardware; the firmware adds a 50 ms delay and a second read.

All three buttons are configured at once from a bit mask:

play_tone/main/main.c — app_main()

    /* GPIOs */
    gpio_config_t out_cfg = {
        .pin_bit_mask = (1ULL << LED_PIN),
        .mode = GPIO_MODE_OUTPUT,
    };
    gpio_config(&out_cfg);
    gpio_set_level(LED_PIN, 0);

    uint64_t btn_mask = 0;
    for (int i = 0; i < NUM_SLOTS; i++)
        btn_mask |= (1ULL << BTN_PIN[i]);

    gpio_config_t btn_cfg = {
        .pin_bit_mask = btn_mask,
        .mode = GPIO_MODE_INPUT,
        .pull_up_en = GPIO_PULLUP_ENABLE,
    };
    gpio_config(&btn_cfg);

Why the amplifier output is bridged

The HT6872 amplifies the difference between the codec’s two outputs rather than one output against ground, which doubles the voltage swing available from a 5 V supply. If both channels carry the same signal the difference is zero and the output is silent. The firmware sets I2S to mono on the left slot, which leaves ROUT1 at zero.

Bridge-tied output. Driving one channel gives half the maximum bridged power; full power would require inverted signals on both channels.

Audio parameters

16 kHz, 16-bit, mono. Three slots of up to 20 seconds each: 16000 × 2 bytes × 20 s = 640 000 B per slot. With the temporary recording buffer, peak allocation is 2 560 000 B against 8 MiB of PSRAM.

play_tone/main/main.c

#define NUM_SLOTS           3
static const int BTN_PIN[NUM_SLOTS] = { 12, 11, 13 };  /* SW1, SW2, SW3 */

#define LONG_PRESS_MS       1000    /* tempo para ativar gravação */

/* =====================================================================
 * Áudio
 * ===================================================================== */
#define SAMPLE_RATE         16000
#define RECORD_MAX_SEC      20
#define MAX_SLOT_BYTES      (SAMPLE_RATE * 2 * RECORD_MAX_SEC)  /* ~640 KB */
#define CHUNK_BYTES         4096

PSRAM is volatile: the three slots are lost when the board loses power. The firmware does not write audio to flash.